The Power System Fault Diagnosis Based On Gray Relational Analysis And Improved Inference Chain

With the enlargement of power system scale, the strcture of power system is more comoplex. When faults occur in power systems, a lager number of alarms are sent to the dispatch center, which give dispathers great difficultly to quick identity the faulty components. The existing fault diagnosis method for power system which based on the digital information can correctly identify the faultly components when protections and circuit breakers operate correctly and no alarm information is missed or misinformation. But when the protections and circuit breakers misoperation and the alarm information missed or misinformation or more protections and circuit breakers information is incorrect and incomplete, the fault diagnosis whiche based on digital information can’t judge the faultly components. Electric quantity information have incomparable advatages than digital information in terms of accuracy,completenessand tolerance, etc, so this paper fuse the electril quantity and digital information, by the method of gray relational analysis and improved inference chain to improve the accuracy of the fault diagnosis.Based on the fault recording data, this paper presents a fault diagnosis method which based on gray relational analysis. For every suspected fault line in power cut area, according to the current sampling value before and after the fault, calculate its three kinds overtical gray correlation of original characteristics sequence, amplitude, energy; by selecting one active reference line, according to current sampling after the fault, calculate three kinds of transverse gray correlation between reference line and suspected fault line. Then weighted fusion vertical gray correlation and transverse gray correlation, we get the comprehensive gray correlation, based on the fault criterion, finding the fault elements. By simulation on the IEEE 14-bus, proving that wherever the fault is in and whatever kinds the fault is belongs to this method can accurately find the fault element.This paper also studied the Bayesian network power systeme fault diagnosis which based on the cavsality inference chain. Due to the large numbers of inference chains, the more complex when structure the inference chain and no considering the missed and misinformation of protections and circuit breakers of the existingpower systeme fault diagnosis which based on inference chian, this paper established a original inference chain for every suspected fault component, by correcting misinformation alarm information and supplementary missed alarm information, correctiong the original inference chain. By evaluation the operation credibility of the protections and circuit breakers, the protections and circuit breakers’s operation information in the inference chain from the binary 1 or 0 fuzzy into [0,1]. Finally, for each inference chain of the suspected fault component creat a Bayesian network model of corresponding model, by the reverse reasoning of Bayesian, wo can get the fault probility. Through the fault of a typical example of the power system, the result supports the method is effective.Finally, this paper fuse the fault degree which base on gray relational analysis with fault probability which based on Bayesian network power systeme fault diagnosis by the improved D-S fusion algorithm, then identify the faulty compenent by the fusion results. Through the parctial examples, the fusion results verified the improved D-S fusion algorithm has good focus and able to correctly identify the faulty compenent.